iBST (version 1.2)

Bagg_Surv: Bagging improper survival trees

Description

Bagging sunbsampling procedure to aggregate several improper trees using either the pseudo-R2 procedure or the adjusted Logrank procedure. Several scores for variables importance are computed.

Usage

Bagg_Surv(xdata, 
     Y.names, 
     P.names, 
     T.names, 
     method = "R2", 
     args.rpart, 
     args.parallel = list(numWorkers = 1), 
     Bag = 100, feat = 5)

Value

A list of ten elements

MaxTreeList

The list of improper survival trees computed during the bagging procedure

IIS

The Index Importance Score

DIIS

The Depth Index Importance Score

DEPTH

The minimum depth importance Score

IND_OOB

A list of length Bag containing the Out Of Bag (OOB) individuals for improper survival tree model

IIND_SAMP

The final list of length Bag of sample individuals used for each improper survival tree

IIND_SAMP

The initial list of sample individuals used for each improper survival tree at teh beginning

Bag

The number of bagging samples retained at the end of the procedure after removing the trees without leaves

indrpart

a vector of TRUE or FALSE with the value FALSE when the corresponding tree is removed from the final bagged predictor

Timediff

The ellapsed time of the Bagging procedure

Arguments

xdata

The learning data frame

Y.names

A vector of the names of the two variables of interest (the time-to-event is follow by the event indicator)

P.names

The names of independant variables acting on the non-susceptible population (the plateau)

T.names

The names of independant variables acting on the survival of the susceptible population

method

The choosen method (either "LR" for the Logrank or "R2" for the proposed pseudo-R2 criterion)

args.rpart

The improper survival tree parameters: a list of options that control details of the rpart algorithm. minbucket: the minimum number of observations in any terminal <leaf> node; cp: complexity parameter (Any split that does not decrease the overall lack of fit by a factor of cp is not attempted); maxdepth: the maximum depth of any node of the final tree, with the root node counted as depth 0. ... See rpart.control for further details

args.parallel

a list containing the number of parallel computing arguments: The number of workers, the type of parallelization to achieve, ... see mclapply for further details.

Bag

The number of Bagging samples to consider

feat

The size of features subsample. A full baging when feat is the total number of features.

Author

Cyprien Mbogning and Philippe Broet

Details

For the Bagging procedure, it is mendatory to set maxcompete = 0 and maxsurrogate = 0 within the args.rpart arguments. This will ensured the correct calculation of the importance of variables and also a better computation time.

References

Mbogning, C. and Broet, P. (2016). Bagging survival tree procedure for variable selection and prediction in the presence of nonsusceptible patients. BMC bioinformatics, 17(1), 1.

Duhaze Julianne et al. (2020). A Machine Learning Approach for High-Dimensional Time-to-Event Prediction With Application to Immunogenicity of Biotherapies in the ABIRISK Cohort. Frontiers in Immunology, 11.

See Also

Bagg_pred_Surv

Examples

if (FALSE) {
 data(burn)
 myarg = list(cp = 0, maxcompete = 0, maxsurrogate = 0, maxdepth = 2)
 Y.names = c("T3" ,"D3")
 P.names = 'Z2'
 T.names = c("Z1", paste("Z", 3:11, sep = ''))
 mybag = 40
 feat_samp = length(T.names)
 set.seed(5000)
 
 burn.BagEssai0 <- suppressWarnings(Bagg_Surv(burn, 
    Y.names, 
    P.names, 
    T.names, 
    method = "LR", 
    args.rpart = myarg, 
    args.parallel = list(numWorkers = 1), 
    Bag = mybag, feat = feat_samp))
 
 burn.BagEssai1 <- suppressWarnings(Bagg_Surv(burn, 
    Y.names, 
    P.names, 
    T.names, 
    method = "R2", 
    args.rpart = myarg, 
    args.parallel = list(numWorkers = 1), 
    Bag = mybag, feat = feat_samp))

}